News · Longevity & Aging
Smoking and waist size beat every epigenetic clock tested
An Aging Cell study followed 1,108 middle-aged Finns for up to nine years and found a model built from age, sex, smoking, alcohol and body shape predicted disease better than any aging clock.
Based on a peer-reviewed cohort study in Aging Cell
- Researchers compared the most commonly used epigenetic clocks against traditional risk factors in 1,108 people aged 34 to 49 at baseline, publishing in Aging Cell.
- The outcome was incidence of age-associated non-communicable chronic disease over 7 to 9 years of follow-up.
- The traditional model used six inputs: age, sex, smoking, alcohol consumption, waist-hip ratio and body mass index.
- A model combining those traditional factors outperformed any model that included an epigenetic clock.
- The authors argue the added value of clocks over simple, affordable risk factors should be established before they are used clinically or for personal health monitoring.
- This tests prediction, not biology. It does not show that clocks fail to measure aging, only that they did not beat a tape measure and a smoking history here.
- One cohort, one age band, and 1,108 people. A larger or older sample could shift the comparison.
The biological-age industry rests on a claim that sounds obviously true: a molecular readout of how your body is aging must tell you more than a questionnaire. A study in Aging Cell put that to a direct test, and the questionnaire won.
The researchers compared the most commonly used epigenetic clocks and traditional risk factors as predictors of incidence of ageing-associated non-communicable chronic disease. The traditional factors were the ones a clinic collects in five minutes.
The comparison nobody had run
Epigenetic clocks predict illness and death. That much is established, repeatedly, and it is why they attracted money and attention.
The question that had been skipped is narrower. As the authors put it, current literature is lacking a formal analysis of the increased prediction accuracy, or the added value, of the epigenetic clocks over traditional risk factors, such as body composition, smoking, or alcohol consumption, in predicting common chronic diseases.
Added value is the phrase doing the work. A marker can be genuinely informative on its own and still contribute nothing once you already know someone’s age and whether they smoke. Nobody had checked which of those was true.
What was measured
The cohort was young enough that disease was still arriving rather than already present. The comparison ran in a 7-to-9-year follow-up in a middle-aged population cohort of 1,108 people, aged 34 to 49 years at baseline.
The traditional side used six inputs, and none of them requires a laboratory: age, sex, smoking, alcohol consumption, waist-hip ratio and body mass index. Against that sat the clocks, each requiring a methylation array and a specialist pipeline.
The result
It was not close in the direction the field expected.
In this cohort, a statistical model consisting of a combination of traditional risk factors outperforms any model including an epigenetic clock.
Any model. Not the average clock, not most clocks. The six cheap variables beat every combination that brought a clock along.
Why the cheap inputs are so hard to beat
Age is the strongest single predictor of most chronic disease, and it is known exactly, with no measurement error at all. Smoking has a large effect that has been characterized for seventy years. Waist-hip ratio and body mass index capture metabolic risk for the price of a tape measure.
A new biomarker does not have to be informative to justify itself. It has to be informative about something that stack does not already cover. That is a much higher bar, and it is the bar this study set.
What the study does not say
It does not say epigenetic clocks fail to measure aging. Methylation changes with age in patterned, reproducible ways, and clocks capture that. The finding is about prediction, not biology, and those come apart routinely.
It also runs on one cohort in one age band. People aged 34 to 49 are early in the disease curve, and a clock might separate people more usefully at 70 than at 40, when more damage has accumulated and methylation patterns have diverged further. Seven to nine years is short for chronic disease.
Where that leaves the tests you can buy
The authors’ conclusion is aimed squarely at that market: the added value of epigenetic clock measurements over simple and affordable traditional risk factors should be clearly established if epigenetic clocks are to be used in clinical settings or as tools of personal health monitoring.
Read carefully, that is not a claim the tests are wrong. It is a claim that nobody has shown they add anything, and that the burden sits with the people selling them.
Which is worth holding onto next time a report returns a biological age three years below your real one. The number may well be measuring something. Whether it is telling you anything your bathroom scales and your smoking history did not is a separate question, and on this evidence it has not been answered.
People also ask
What is an epigenetic clock?
A statistical model that estimates biological age from DNA methylation, the chemical tags that sit on DNA and change in patterned ways across a lifetime. Feed in a methylation profile and the clock returns an age, which can differ from the number of birthdays. The best-known versions include Horvath, Hannum, PhenoAge, GrimAge and DunedinPACE, and they underpin most of the consumer biological-age tests now sold.
So do epigenetic clocks not work?
That is not what this shows, and the distinction matters. Clocks do predict illness and death, which is well established and is why they got attention. The question this study asks is narrower and more practical: do they predict better than information a clinic already has for free? Here they did not. A clock can be measuring something real about aging and still add nothing to a prediction that already includes your age, your smoking history and your waist.
Why do age, sex and smoking do so well?
Because they carry an enormous amount of information about disease risk and they are measured without error. Age alone is the single strongest predictor of most chronic disease. Smoking has a large, well-characterised effect. Waist-hip ratio and body mass index capture metabolic risk cheaply. A biomarker has to beat that stack to earn its cost, and beating it is a high bar.
Does this apply to the biological-age test I can buy?
It is directly relevant, and it is the reason the authors raise personal health monitoring specifically. A consumer test reports a number that feels informative. This study asks whether that number tells you anything your existing risk factors did not already say, and in this cohort the answer was no. That does not make the test wrong; it makes its added value unproven, which is a different and more useful criticism.
What are the study's limits?
It is one cohort of 1,108 Finns who were 34 to 49 at baseline, followed for 7 to 9 years. That age band matters: clocks may perform differently in older people, where more disease accumulates and methylation patterns diverge further. The follow-up is also short for chronic disease. A larger study in an older population could reach a different comparison.
References
- Kostiniuk D, Szekely F, Lyytikainen LP, et al. Traditional Disease Risk Factors Outperform Epigenetic Clocks as Predictors of Non-Communicable Disease Morbidity in a Middle-Aged Cohort. Aging Cell (2026).
- National Institute on Aging. Biomarkers of Aging.
- Centers for Disease Control and Prevention. Assessing Your Weight.